OctaneRender™ Standalone v1.0 beta3.02x [experimental]

[pixelrush]

Yeah well apparently its much nicer now but I wonder if there is something amiss in there. All the other figures seem to have held up except that one, still you know what you are about, I suppose

[abstrax]

One more thing: PMC has a warmup phase of 15 samples/pixel, then the speed jumps up and then decreases a bit until it settles at some rate. Because of that, any PMC measurements should be done at least after 100 samples/pixel.

Cheers,
Marcus

[abstrax]

Yeah well apparently its much nicer now but I wonder if there is something amiss in there all the other figures seem to have held up except that one, still you know what you are doing, I suppose

Just by looking at the renderings you see that they are fairly different. The problem is that Fresnel reflection is now disabled when the index in the glossy material is <= 1.

Cheers,
Marcus

[pixelrush]

OK still seems like an overly large impact to me but...

[bepeg4d]

Thanks for the feedback. I think there is something fishy going on on Mac OS. Is the 560Ti officially supported on Mac OS? Which CUDA driver does support it at all? I think the CUDA driver is the problem here, but that's just a wild guess.

Cheers,
Marcus

tomorrow i'll try to switch to 5.0.17 and 5.0.24 to see if there are differences
ciao beppe

[bepeg4d]

it must to be said that from os x 10.7.5, and i suspect from 10.8 and above, it's no more possible to come back to cuda 4.x, so the only possibility are 5.0.17/24/36.
with 5.0.17 no card are detected by octane, even the gt120 (standard mac-nvidia card)
with 5.0.24 there are no differences with 5.0.36
i wonder, if it's a cuda driver problem, why the 2.59 and 3.01 are working as expected and only the new builds from 3.02 to 3.03 are slower?
could it be the toolkit that must to be uninstalled?
if yes i must to figure out how to remove it
ciao beppe

[abstrax]

it must to be said that from os x 10.7.5, and i suspect from 10.8 and above, it's no more possible to come back to cuda 4.x, so the only possibility are 5.0.17/24/36.
with 5.0.17 no card are detected by octane, even the gt120 (standard mac-nvidia card)
with 5.0.24 there are no differences with 5.0.36
i wonder, if it's a cuda driver problem, why the 2.59 and 3.01 are working as expected and only the new builds from 3.02 to 3.03 are slower?
could it be the toolkit that must to be uninstalled?
if yes i must to figure out how to remove it
ciao beppe

Beta 2.59 and beta 3.01 were built using CUDA toolkit 4.0, while beta 3.02/3.02 were built using CUDA 4.2. Somehow this doesn't seem to go well with the CUDA driver. I see a few possibilities we can try and on Monday I will contact you via PM to try to sort this issue out.

Thanks for your help,
Marcus

[Vanya]

Hi. This is my experiment in 3.3 version.

[matej]

Updated benchmark, now with the latest stable Linux drivers 304.51

With the new drivers, compared to older 295.75:

* Octane 259 gains in speed just slightly in some cases
* Octane 301 gains in speed quite noticeably
* Octane 302 x1 is exactly the same as with the previous drivers
* Octane 302 x2 is exactly the same as with the previous drivers

I didn't benchmark 303 last time

[gzavye]

My results.
Faster results at the left, slower at right
https://docs.google.com/spreadsheet/ccc ... xRkE#gid=0

UPDATE:
I've added 2.59 and 3.03
2.59 is the faster except for screws. In 3.XX series 3.02X1 is the faster version.